1. Field of the Invention
The present invention relates to an electron-beam lithography system, or more particularly, to an electron-beam lithography system for drawing patterns using an electron beam by continuously moving a stage.
2. Description of the Related Art
With advancement in a microprocessing technology, the trend of semiconductor integrated circuits is toward very dense integrated circuits. The performance which the microprocessing technology is required to offer must be severely evaluated. Among lithography technologies, the photolithography technology implemented in a step-and-repeat photolithography system with demagnification or the like in the past is expected to reach its limits in the near future. An electron-beam lithography technology is expected to take over from photolithography technology and be used for a new generation of microprocessing.
The drawbacks of the electron-beam lithography technology, been pointed out in the past, are low processing speed and poor manufacturing efficiency. For conventional electron-beam lithography systems, similarly to the step-and-repeat photolithography system with demagnification, a step-and-repeat photolithography method with demagnification has been adopted. According to this method, a sample (wafer) is placed on a stage. After a predetermined area is exposed and thus patterned, the stage is moved in order to expose a subsequent area for patterning. This procedure is repeated, whereby the whole sample is exposed and thus patterned. For example, one wafer is exposed to produce 60 chips each having a width of 15 mm. In this case, since an area in which an electron beam can be deflected is approximately 1.5 mm in width, the stage must be moved 100 times in order to produce each chip. For a whole wafer, the stage must be moved 6000 times. Exposure cannot be carried out while the stage is moved or until an irradiated position to which an electron beam is irradiated and which is changed with movement of the stage is determined accurately. This time interval shall be referred to as a stage settlement wait time herein. Movements of the stage are mechanical. For precise movement, therefore, it takes some time. The stage settlement time is therefore considerably long, or at present, about 0.5 sec. The time required for moving the stage 6000 times is as much as 50 min. There is therefore a difficulty in improving the processing speed.
Development of various technologies is under way in efforts to improve a throughput. A continuous movement lithography (stage scan) method is one example of these technologies. According to this method, the stage is moved continuously for exposure. FIG. 1A and FIG. 1B are explanatory diagrams concerning the continuous movement lithography method. FIG. 1A is concerned with a method of exposing a wafer with the stage being moved in one direction, while FIG. 1B is concerned with a method of exposing the wafer with the stage being advanced and returned in opposite directions.
As illustrated, a plurality of chips 90 is produced from a sample (wafer) 18. Only four chips are shown in the drawings. Normally, several tens to several hundreds of chips are produced. As mentioned above, an area in which an electron beam can be deflected is smaller than each chip 90. Each chip 90 is therefore segmented into a plurality of areas 91 and then exposed. According to the step-and-repeat photolithography method with demagnification, when the stage is moved to the center of each area, the area is exposed. In contrast, according to the continuous movement lithography method, as illustrated, the areas 91 that are aligned on the same column in the same column of chips 90 are exposed by moving the stage. In other words, an exposed area is segmented into a plurality of rectangular areas extending over the borders among chips. The stage is moved continuously over the rectangular areas. According to this method, the stage settlement wait time can be nullified and the processing speed can be improved. For example, assuming that the area in which an electron beam can be deflected is approximately 1.5 mm wide, an area having dimensions of 1.5 mmxc3x97ten several centimeters can be patterned during a single scan without the necessity of taking a stage settlement wait time. The total stage settlement wait time is therefore decreased to about {fraction (1/100)}. The processing speed therefore improves greatly. For methods of scanning a wafer with an electron beam to pattern it, the continuous movement lithography method can be said to be an essential technology. According to the continuous movement lithography method, a sample moves during patterning. The position of an electron beam must therefore be corrected accordingly. Controlling the position of an electron beam is therefore more complex than that according to the step-and-repeat photolithography method with demagnification. The continuous movement lithography method falls into the method of exposing a wafer with a stage being moved in one direction as shown in FIG. 1A, and the method of exposing a wafer with the stage being advanced and returned in opposite directions as shown in FIG. 1B. The method shown in FIG. 1B offers a higher processing speed because it is unnecessary to return to one extreme end. However, since backlash and the like pose a problem, a high-precision moving mechanism is needed.
FIG. 2 shows the configuration of an electron-beam lithography system for performing lithography according to the continuous movement lithography method. In FIG. 2, there are shown a processor 1, a magnetic disk 2, and a magnetic tape drive 3. These units are interconnected over a bus 4, and connected to a data memory 6 and stage control circuit 7 via an interface circuit 5 over the bus 4.
A housing (column) 8 accommodates an electron gun 9, a lens 10, a blanking electrode 11, a lens 12, a feedback coil 13, a sub-deflector coil 14, a lens 15, a main deflector coil 16, a stage 17, and a sample 18. The sample (wafer) 18 is placed on the stage 17. The stage 17 is moved in the X and Y directions according to an output signal of the stage control circuit 7.
Moreover, data read from the data memory 6 is supplied to a pattern correction circuit 20 via a pattern generation circuit 19. The pattern correction circuit 20 applies a blanking signal to the blanking electrode 11 via an amplifier 21. Moreover, the pattern correction circuit 20 applies a signal to the coils 13, 14, and 16 via D/A converters (DAC) 22, 24, and 26 and amplifiers 23, 25, and 27.
An electron beam radiated from the electron gun 9 passes through the lens 10. The electron beam is then transmitted or intercepted by the blanking electrode 11, and reshaped into a rectangular beam of parallel rays having any spot size of, for example, 3 xcexcm or less. Thereafter, the feedback coil 13, sub-deflector coil 14, and main deflector coil 16 deflect the electron beam. The electron beam is then converged on the surface of the sample through the projection lens 15. Areas where the feedback coil 13, sub-deflector coil 14, and main deflector coil 16 can deflect the beam get larger in that order. Specifically, the area where the feedback coil 13 can deflect the beam is smaller than that where the sub-deflector coil 14 can. The area where the sub-deflector coil 14 can deflect the beam is smaller than that where the main deflector coil 16 can. For ensuring a large area where the beam can be deflected, the number of windings of a coil must be increased accordingly. The response speeds of the coils get lower in reverse order. In other words, a settlement wait time required by the feedback coil 13 is the shortest. Settlement wait times required by the sub-deflector coil 14 and main deflector coil 16 get longer in that order. In the foregoing configuration, the exposed areas 91 shown in FIG. 1A and FIG. 1B correspond to the area where the main deflector coil 16 can deflect the beam.
Referring back to FIG. 1A and FIG. 1B, whichever of the method of exposing a wafer with a stage being moved in one direction and the method of exposing it with the stage being advanced and returned in opposite directions is adopted, a position detection mark is inscribed near each chip 90. The position detection mark is detected with an electron beam, whereby an exposed position is determined. For producing IC chips, patterning must be performed on several layers. The stage is positioned relative to the position detection mark, whereby patterns drawn on the layers are aligned with one another. In reality, before patterning is performed, the position detection mark near each chip is detected in order to calculate a positional deviation of the chip or a misalignment between chips. When each chip is exposed to define a stripe thereon, the positional deviation is corrected.
In the electron-beam lithography system, the resolution offered with an electron beam and the precision in positioning are so high that very fine patterns can be drawn. However, a phenomenon that an electron beam drifts to change its position takes place because members including deflectors are charged up and a chamber forming the column and optical members expands thermally. This becomes a factor degrading the precision of an irradiated position at which an electron beam is irradiated. For preventing the phenomenon, the members including deflectors are made of a material that is hardly charged up or structured not to be charged up readily. Moreover, materials exhibiting very small coefficients of thermal expansion are used, and other measures are taken. Consequently, a beam drift has become much smaller.
If the line width of patterns is not very thin, the influence of the beam drift does not pose a serious problem. However, when finer patterns must be drawn, the influence of the beam drift cannot be ignored. The position of a beam is therefore measured at regular intervals in order to calculate the magnitude of a positional deviation, that is, a drift. The position of the beam is corrected by the calculated magnitude, and exposure is then carried out.
FIG. 3 is an explanatory diagram concerning a conventional method of correcting a drift. FIG. 4 is a flowchart describing drift correction based on the conventional drift correcting method. As shown in FIG. 3, according to a related art, when each column of chips has been exposed for defining all stripes, a position detection mark is detected to calculate a positional deviation of one chip from another. Before it starting to exposure the next column of chips so as to define stripes, a correction value for canceling the positional deviation is re-set. A solid line indicates a beam drift, dots indicate measured points and measured beam drifts, and small circles indicate correction values. For correcting the beam drift as graphically shown in FIG. 3, as described in FIG. 4, an initial position of a beam is measured and initialization is carried out at step 100. Stripes are defined at step 101. It is judged at step 102 whether defining stripes on a column of chips has been completed. If it is not completed, control is returned to step 101 and stripes are defined on the next column of chips. If defining stripes on the column of chips has been completed, it is judged at step 103 whether defining stripes on all columns of chips has been completed. If it is completed, exposure is terminated. If not, a position detection mark (reference mark) is detected to calculate a beam drift (a positional deviation) ex at step 104. A correction value is set to xe2x88x92ex at step 105. Control is then returned to step 101.
As shown in FIG. 3 and FIG. 4, a beam drift has been corrected for each column of chips in the past. The reasons why this correction has been adopted will be described below.
Occurrence of a beam drift is uncertain and a beam drift is hard to predict. Beam drifts occurring when layers are exposed are not the same in magnitude. The beam drift occurring on each layer is measured as a deviation of an exposed position for the layer. What counts is the absolute value of a deviation of one layer from another or a misalignment between layers. The absolute value corresponds to a difference between a value pointed out with a small circle in FIG. 3 and an associated value on the solid line indicating the beam drift.
In contrast, if a beam drift occurs while the same chip on the same layer is exposed to define adjoining stripes wiring extending over the stripes has mismatched portions. FIG. 5A to FIG. 5C are explanatory diagrams concerning occurrence of a misalignment between stripes on the same layer within the same chip. If a beam drift varies as indicated with a solid line in FIG. 5A, and if correction values indicated with small circles are adopted, deviations of exposed positions for defining stripes a, b, and c are indicated as A, B, and C. The sum of the deviations corresponds to a deviation of the layer from an adjoining one. The stripes a and b have therefore, as shown in FIG. 5B, a misalignment P between them. Herein, a difference in beam drift between adjoining stripes shall be small enough. If the difference were large, exposure itself would not be able to be carried out.
When a deviation of an exposed position for stripe c is large, and before exposure is performed to define the next stripe d, a correction value is modified by E corresponding to C. A deviation of an exposed position for stripe d is, as shown in FIG. 5C, D. Since the deviation for stripe c is C, a difference in deviation between stripes c and d is Cxe2x88x92D=Q. The difference thus increases abruptly. When a beam drift is measured in the course of exposing the same column of chips, if an exposed position is corrected by the measured beam drift, a misalignment between adjoining strips increases abruptly.
In general, a tolerance of a misalignment between layers is larger than that of a misalignment between adjoining stripes on the same layer within the same chip. In the related art, therefore, every time exposing a column of chips is completed, a position detection mark is detected and a beam drift is measured and corrected. While each column of chips is exposed, a beam drift is regarded not to vary but remain constant. As shown in FIG. 3, assume that adjoining stripes are concerned and belong to different columns of chips. In this case, even if the correction value is modified, since the stripes do not exist on the same chip, the tolerance of a misalignment between the stripes is set to the one between layers.
However, the area of a chip has increased along with the trend of ICs towards more complex structures. ICs whose chips must each be exposed for defining ten several stripes have made their debuts. It therefore takes much time to expose one column of chips for defining all stripes. A variation in beam drift occurring meanwhile is therefore so large as to exceed the tolerance of a misalignment between layers. This poses a problem. For solving the problem, it is necessary to measure and correct a beam drift during exposure of one column of chips. When the correction is carried out, the misalignment between adjoining stripes increases. The correction, therefore, cannot be carried out.
Moreover, before a beam drift gets too large, after every exposure is completed for defining each stripe, the beam drift may be measured and corrected. However, this leads to a decrease in throughput. For measuring a beam drift, for example, an alignment mark on a wafer is regarded as a reference mark, the reference mark is scanned with an electron beam, and reflected electrons are detected. A time required for the electron beam to cross the reference mark is thus measured. For measuring a beam drift, therefore, the stage must be moved so that an electron beam will be positioned at the reference mark. Exposure must therefore be suspended in order to move the stage. Consequently, the throughput of the system decreases. If the beam drift is measured and corrected before a variation in beam drive becomes very large, it means that the beam drift is measured frequently. This also leads to a decrease in throughput. As mentioned above, even if a beam drift is large in magnitude, when the frequency of correcting the beam drift is raised, an exposed position can be set precisely, though the throughput decreases. When the time interval between corrections is extended, if a beam drift gets larger, precision is degraded impermissibly. The frequency of correcting a beam drift and the throughput of the system, and the time interval between corrections and the precision have a relationship of trade-off.
The first object of the present invention is to make it possible to correct a beam drift even during exposure of the same column of chips so that the correction can be performed frequently. The second object of the present invention is to minimize the frequency of correcting a beam drift without causing intolerable degradation in precision.
FIG. 6A and FIG. 6B are explanatory diagrams concerning the principles of the present invention.
As shown in FIG. 6A, in an electron-beam lithography method and electron-beam lithography system according to the first aspect of the present invention, a beam drift is measured frequently. When a beam drift is measured during exposure of each column of chips, a positional deviation is not corrected at a time but corrected step by step while a plurality of stripes is defined. Consequently, it is prevented that a misalignment between adjoining strips increases.
Moreover, in an electron-beam lithography method and electron-beam lithography system according to the second aspect of the present invention, a variation in beam drift is measured from a plurality of measured beam drifts. As shown in FIG. 6B, when a beam drift is small, an interval between measurements is extended. As shown in FIG. 6C, when the beam drift is large, the interval between measurements is shortened.
The electron-beam lithography method in accordance with the present invention is a method of patterning a sample in units of a stripe using an electron beam. The electron-beam lithography method consists of a step of measuring a drift made by an electron beam by detecting a reference mark, and a step of correcting the drift. According to the first aspect of the present invention, at the step of correcting a drift included in the lithography method, a drift is corrected step by step while a plurality of strips is defined after the measurement of the drift. According to the second aspect of the present invention, the lithography method further includes a step of calculating a drift per unit time and a step of calculating a time elapsing until the next step of detecting a drift according to the drift per unit time. At the step of calculating the drift per unit time, the drift per unit time is calculated from an elapsed time between a plurality of steps of detecting a drift and a difference between measured drifts.
Moreover, the electron-beam lithography system in accordance with the present invention consists mainly of an electron gun, a deflecting means for deflecting an electron beam emitted from the electron gun while converging it on a sample, and a moving mechanism to be moved with the sample placed thereon. While the sample is moved using the moving mechanism, the sample is exposed to the electron beam in order to define stripes having a predetermined width. The sample is thus patterned. This procedure is repeated. The electron-beam lithography system further includes a drift detecting means for detecting a reference mark and measuring a drift made by the electron beam, and a drift correcting means for correcting a drift. According to the first aspect of the present invention, the drift correcting means corrects a drift step by step at the time of defining a plurality of stripes succeeding the time of measuring a drift. According to the second aspect of the present invention, the electron-beam lithography system further includes a means for calculating a drift per unit time and a means for calculating a time elapsing until the next measurement of a drift according the drift per unit time. The means for calculating the drift per unit time calculates the magnitude of a drift per unit time using an elapsed time necessary for the drift detecting means to measure a drift, and a difference between measured drifts.
The constituent features in accordance with the second aspect can be added to the electron-beam lithography method and system in accordance with the first aspect.
According to the first aspect of the present invention, as shown in FIG. 6A, a reference mark may be detected during patterning of chips, and a beam drift d1 may be measured. In this case, the beam drift d1 is corrected step by step at the subsequent times of defining a plurality of stripes (four stripes in the drawing) indicated with small triangles. It can therefore be prevented that a misalignment between adjoining stripes increases abruptly. Even during patterning of chips, a relatively large beam drift can be corrected.
According to the second aspect of the present invention, as shown in FIG. 6B, when a beam drift is small, the number of stripes to be defined until the next measurement of a beam drift is increased (to six in the drawing). As shown in FIG. 6C, when a beam drift is large, the number of stripes to be defined until the next measurement of a beam drift is decreased (to two in the drawing). Thus, an interval between measurements and corrections of a drift is shortened according to the magnitude of a beam drift. Even when a beam drift is large, precision can be ensured for an exposed position. When a beam drift is small, even if the frequency of measurement is decreased, precision can be ensured for an exposed position. Therefore, the interval between measurements is extended in order to prevent a decrease in throughput.
In a method according to both the first and second aspects of the present invention, at a step of correcting a drift, the number of stripes, N, to be defined until the next step of measuring a drift is calculated from a calculated time elapsing until the next step of measuring a drift and an exposure time required for defining each stripe. A difference between a drift predicted to be measured at the next step of measuring a drift and a current drift correction value is calculated as dx. The drift is then corrected in units of dx/N while N stripes are defined.
Furthermore, an upper limit ux is determined for a difference in correction value between stripes. If the calculated dx/N is equal to or larger than the upper limit ux, a misalignment between the stripes becomes large. It is therefore preferred that an alarm is made in this case. In response to the alarm, measurements may be taken. Alternatively, a position exposed when the alarm is made is stored for future review of whether the misalignment is permissible for the position.
A standard value ssx may be determined in advance for a difference in correction values between stripes. If the calculated dx/N is larger than or equal to the absolute value of the standard value ssx, dx/N is set to the standard value ssx. A correction value is modified by the standard value ssx for defining each stripe. After a predicted temporal variation in beam drift is canceled out, a value canceling out the temporal variation in beam drift may be adopted as a correction value.
According to the first aspect, when a drift is detected during exposure of the same column of chips, the drift is corrected in units of the aforesaid dx/N. When a drift is detected at the completion of exposure of the same column of chips, the drift should preferably be corrected at one time.
According to the second aspect, the time elapsing until a drift is measured and corrected next may be calculated as a measurement interval elapsing until the next step of measuring a drift tm=sw/vx. between successive detections of a drift. Herein, the standard value sx is determined in advance at the step of measuring a drift next. At the step of calculating the number of stripes, the measurement interval tm=sw/vx is calculated from the drift per unit time, vx, and the standard value of a drift, sx.
Preferably, an upper limit T should be determined for the time elapsing between the steps of measuring a drift. When the measurement interval tm exceeds the upper limit T, the upper limit T should preferably be adopted as the time elapsing until the next step of measuring a drive. According to the aforesaid method, when a variation in drift detected through two measurements is limited, measurement of a drift may be carried out for a long time. The variation may presumably increase thereafter. It is therefore not preferred that measurement of a drift is not carried out for a very long time. Adopting the constituent features of the present invention can prevent this incident.